[ David Greene wrote in 121124 MMDigest:

> I have three goals for this project:
>   - Allow computer-controlled playback via MIDI files or similar;
>   - Provide accurate electronic recording;
>   - Create new rolls from digital/MIDI sources.

After reading David Greene's post in MMD 121124, I tried to focus on
something positive to say.  His enthusiasm is commendable and I do not
wish to dampen that.  Part of his planned project would be easy to
complete.  Other parts have kept people busy for years with limited
success.

The easy part is converting his piano to play from MIDI files.  There
are numerous resources covering that topic in the MMD archives and
elsewhere, so I'll not comment.  But equipping a piano to record
keystrokes reliably is no small task.

Attempted recordings of keystrokes with dynamics kept the entire piano
roll industry busy for a quarter of a century during the heyday years
-- it was the holy grail.  Real success occurred through the efforts
of Dr. Clarence Hickman at Ampico.  Ampico began using his methods early
in 1926 for production of Ampico A rolls.  That was soon followed by
using the same technology for the newer Ampico B system with those
rolls appearing late in 1927.

Most modern electronic keyboards contain technology for determining
note intensity, which in turn is assigned a MIDI velocity/intensity
value as part of the data for each note "on" event.  Using the sensors
from one of these keyboards would be a logical starting point for
adapting your Chickering.  However, the sensors on most of these
keyboards render only coarse generalizations of real hammer velocity.
There are no hammers in electronic keyboards and so the velocity number
is somewhat arbitrary.  Using this system on an Ampico to record an
actual roll performance is destined to mediocrity at best.

As far as the fuss about optical scanning of rolls, that really isn't
a problem as many scanned files are already available.  If these files
are played back through a proper MIDI roll interface using the native
expression functions within the piano, the results should be
satisfactory.

Scanned files of reproducing piano rolls that have been converted to
play MIDI devices is another story.  Most programs that emulate the
expression systems of reproducing pianos are simple approximations.
Very important characteristics of the pneumatic system are seldom
understood by the persons writing the translations, and thus the
results lack nuance and often sound like elevator music.

Finally, the roll creation desire expressed by David is both easily
attained and formidable.  Fine perforator operations are maintained
by several MMDers.  Sending a MIDI file to them, of keystroke
information, will result in crisply cut rolls in a short period of
time.  Putting any satisfactory form of reproducing code on those rolls
is a task of huge proportions.  Coding has never been automated with
any success.  The only system that could be easily automated would be
the Duo-Art.  Human editing is always required.

So my advice to David is limit the scope of your project by evaluating
what is feasible, then research in detail your new endeavor.  Part of
the research should include what has already been done in your selected
field.  Study and learn the Ampico code, which is much more complex
than most people think.

Bob Taylor
Missouri